SurfAI: Advanced Browser Automation Framework

A powerful Rust-based browser automation framework with human-like behavior simulation and stealth capabilities. Built on top of chromiumoxide with advanced scripting, element interaction, and MCP (Model Context Protocol) server integration.

Features

🚀 Core Capabilities

• Stealth Browser Management: Anti-detection capabilities with realistic browser fingerprinting
• Human-like Interactions: Natural mouse movements, typing patterns, and browsing behavior
• Advanced Element Detection: Intelligent element labeling and interaction systems
• CDP Script Execution: Custom Chrome DevTools Protocol script execution
• Page Scraping: Comprehensive data extraction from web pages
• MCP Server Integration: Model Context Protocol server for AI agent integration

🛡️ Stealth Features

• User agent spoofing and browser fingerprint masking
• Realistic timing patterns and random delays
• CAPTCHA detection and avoidance
• Session management with cookies and browsing history simulation
• Mouse movement with natural curves and variations

🎯 Element Interaction

• Smart element detection and labeling
• Google Search automation with human-like patterns
• Form filling and button clicking
• Screenshot capture and visual debugging
• DOM manipulation and event simulation

Installation

Add this to your Cargo.toml:

[dependencies]
surfai = "0.1.2"
tokio = { version = "1.0", features = ["full"] }
serde = { version = "1.0", features = ["derive"] }
serde_json = "1.0"

Quick Start

Basic Browser Automation

use surfai::*;

#[tokio::main]
async fn main() -> Result> {
    // Create a new browser instance
    let browser = BrowserManager::new(false).await?;
    let page = browser.new_page("https://google.com").await?;

    // Initialize components
    let scraper = PageScraper::new(page.clone());
    let interactor = ElementInteractor::new(page.clone());

    // Scrape page data
    let page_data = scraper.scrape_page().await?;
    println!("Page title: {}", page_data.title);

    // Find interactive elements
    let elements = interactor.find_interactable_elements().await?;
    println!("Found {} interactive elements", elements.len());

    // Take screenshot
    let screenshot = scraper.take_screenshot().await?;
    std::fs::write("screenshot.png", screenshot)?;

    browser.close().await?;
    Ok(())
}

Human-like Google Search

use surfai::*;

#[tokio::main]
async fn main() -> Result> {
    // Create stealth browser
    let browser = BrowserManager::new_stealth_browser(false).await?;
    let page = browser.new_page("https://google.com").await?;

    let human = HumanInteractor::new(page.clone());

    // Apply stealth settings
    human.apply_stealth().await?;
    human.setup_session().await?;

    // Perform human-like search
    let search_query = "rust programming language";
    human.search_with_realistic_pattern(search_query).await?;
    human.submit_search().await?;

    // Wait for results and extract
    human.wait_for_results(30).await?;
    let results = human.extract_search_results().await?;

    println!("Found {} results", results.len());
    for (i, result) in results.iter().enumerate().take(5) {
        println!("{}. {}", i + 1, result.get("title").unwrap_or(&"No title".to_string()));
        println!("   Link: {}", result.get("link").unwrap_or(&"No link".to_string()));
    }

    browser.close().await?;
    Ok(())
}

MCP Server Integration

The framework includes a built-in MCP (Model Context Protocol) server for seamless AI agent integration.

Starting the MCP Server

use surfai::mcp::MCPBrowserServer;

#[tokio::main]
async fn main() -> Result> {
    let server = MCPBrowserServer::new();

    // Handle MCP requests
    let request = MCPRequest {
        id: "1".to_string(),
        method: "tools/list".to_string(),
        params: serde_json::json!({}),
    };

    let response = server.handle_request(request).await;
    println!("MCP Response: {:?}", response);

    Ok(())
}

Available MCP Tools

The MCP server provides the following tools for AI agents:

1. create_browser_session

Creates a new browser session with optional stealth mode.

Parameters:

• headless (boolean, default: true): Run browser in headless mode
• stealth (boolean, default: false): Enable stealth anti-detection features

Example:

{
  "name": "create_browser_session",
  "arguments": {
    "headless": false,
    "stealth": true
  }
}

2. navigate_to_url

Navigate to a specific URL in an existing session.

Parameters:

• session_id (string, required): Browser session ID
• url (string, required): Target URL

Example:

{
  "name": "navigate_to_url",
  "arguments": {
    "session_id": "uuid-session-id",
    "url": "https://google.com"
  }
}

3. search_google

Perform a Google search with human-like behavior patterns.

Parameters:

• session_id (string, required): Browser session ID
• query (string, required): Search query

Example:

{
  "name": "search_google",
  "arguments": {
    "session_id": "uuid-session-id",
    "query": "rust programming language"
  }
}

4. get_page_elements

Extract all interactable elements from the current page.

Parameters:

• session_id (string, required): Browser session ID
• element_types (array, optional): Specific element types to filter

Example:

{
  "name": "get_page_elements",
  "arguments": {
    "session_id": "uuid-session-id",
    "element_types": ["input", "button", "a"]
  }
}

5. click_element

Click on an element with human-like mouse movements.

Parameters:

• session_id (string, required): Browser session ID
• selector (string, required): CSS selector for the element

Example:

{
  "name": "click_element",
  "arguments": {
    "session_id": "uuid-session-id",
    "selector": "button[type='submit']"
  }
}

6. type_text

Type text into an element with realistic human typing patterns.

Parameters:

• session_id (string, required): Browser session ID
• selector (string, required): CSS selector for the input element
• text (string, required): Text to type

Example:

{
  "name": "type_text",
  "arguments": {
    "session_id": "uuid-session-id",
    "selector": "input[name='q']",
    "text": "Hello world"
  }
}

7. scrape_page

Extract all data from the current page including title, content, links, images, and forms.

Parameters:

• session_id (string, required): Browser session ID

Example:

{
  "name": "scrape_page",
  "arguments": {
    "session_id": "uuid-session-id"
  }
}

8. take_screenshot

Capture a screenshot of the current page.

Parameters:

• session_id (string, required): Browser session ID

Example:

{
  "name": "take_screenshot",
  "arguments": {
    "session_id": "uuid-session-id"
  }
}

9. close_session

Close a browser session and clean up resources.

Parameters:

• session_id (string, required): Browser session ID

Example:

{
  "name": "close_session",
  "arguments": {
    "session_id": "uuid-session-id"
  }
}

MCP Usage Flow

1. Create Session: Use create_browser_session to initialize a browser
2. Navigate: Use navigate_to_url to go to target websites
3. Interact: Use click_element, type_text, or search_google for interactions
4. Extract Data: Use scrape_page, get_page_elements, or take_screenshot for data extraction
5. Cleanup: Use close_session to properly close the browser

MCP Integration Example

use surfai::mcp::*;

async fn ai_agent_workflow() -> Result> {
    let server = MCPBrowserServer::new();

    // Step 1: Create browser session
    let create_request = MCPRequest {
        id: "1".to_string(),
        method: "tools/call".to_string(),
        params: serde_json::json!({
            "name": "create_browser_session",
            "arguments": {
                "headless": false,
                "stealth": true
            }
        }),
    };

    let response = server.handle_request(create_request).await;
    let session_id = response.result.unwrap()["session_id"].as_str().unwrap();

    // Step 2: Perform Google search
    let search_request = MCPRequest {
        id: "2".to_string(),
        method: "tools/call".to_string(),
        params: serde_json::json!({
            "name": "search_google",
            "arguments": {
                "session_id": session_id,
                "query": "AI automation tools"
            }
        }),
    };

    let search_response = server.handle_request(search_request).await;
    println!("Search results: {:?}", search_response.result);

    // Step 3: Take screenshot
    let screenshot_request = MCPRequest {
        id: "3".to_string(),
        method: "tools/call".to_string(),
        params: serde_json::json!({
            "name": "take_screenshot",
            "arguments": {
                "session_id": session_id
            }
        }),
    };

    let screenshot_response = server.handle_request(screenshot_request).await;

    // Step 4: Close session
    let close_request = MCPRequest {
        id: "4".to_string(),
        method: "tools/call".to_string(),
        params: serde_json::json!({
            "name": "close_session",
            "arguments": {
                "session_id": session_id
            }
        }),
    };

    server.handle_request(close_request).await;

    Ok(())
}

API Reference

Core Components

BrowserManager

Main browser management interface.

Methods:

• new(headless: bool) -> BrowserResult: Create standard browser
• new_stealth_browser(headless: bool) -> BrowserResult: Create stealth browser
• new_page(url: &str) -> BrowserResult: Create new page
• close(self) -> BrowserResult: Close browser

HumanInteractor

Human-like interaction simulation.

Key Methods:

• apply_stealth() -> BrowserResult: Apply stealth settings
• type_human(selector: &str, text: &str) -> BrowserResult: Human-like typing
• click_human(selector: &str) -> BrowserResult: Human-like clicking
• search_with_realistic_pattern(query: &str) -> BrowserResult: Realistic search
• extract_search_results() -> BrowserResult>>: Extract results

ElementInteractor

Element detection and interaction.

Key Methods:

• get_all_elements() -> BrowserResult>: Get all elements
• get_interactable_elements() -> BrowserResult>: Get interactive elements
• label_elements(tags: Option>) -> BrowserResult>: Label elements
• click_element(selector: &str) -> BrowserResult: Click element
• type_text(selector: &str, text: &str) -> BrowserResult: Type text

PageScraper

Web page data extraction.

Key Methods:

• scrape_page() -> BrowserResult: Scrape all page data
• take_screenshot() -> BrowserResult>: Take screenshot
• extract_links() -> BrowserResult>: Extract links
• extract_images() -> BrowserResult>: Extract images

Configuration

Stealth Browser Settings

The stealth browser includes comprehensive anti-detection features:

let browser = BrowserManager::new_stealth_browser(false).await?;

Stealth Features:

• User agent spoofing
• Plugin and extension masking
• Canvas fingerprinting protection
• WebGL fingerprinting protection
• Timezone and language spoofing
• Screen resolution normalization

Human Behavior Patterns

Customize human-like behavior:

let human = HumanInteractor::new(page);

// Configure timing patterns
human.random_delay(1000, 2000).await;  // Random delay between 1-2 seconds
human.simulate_reading_delay("Sample text").await;  // Reading-based delay
human.simulate_hesitation().await;  // Simulate user hesitation

Examples

The examples/ directory contains comprehensive usage examples:

• basic_usage.rs: Basic browser automation
• google_search.rs: Google search automation
• human_google_search.rs: Human-like Google search
• label_elements.rs: Element labeling and interaction

Run examples with:

cargo run --example basic_usage
cargo run --example google_search
cargo run --example human_google_search

Error Handling

The framework uses a comprehensive error system:

use surfai::BrowserResult;

match browser.new_page("https://example.com").await {
    Ok(page) => {
        // Success
    }
    Err(BrowserError::Navigation(msg)) => {
        println!("Navigation error: {}", msg);
    }
    Err(BrowserError::ElementNotFound(selector)) => {
        println!("Element not found: {}", selector);
    }
    Err(e) => {
        println!("Other error: {}", e);
    }
}

Requirements

• Rust 1.70+
• Chrome or Chromium browser installed
• tokio runtime
• Network access for web automation

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests for new functionality
5. Submit a pull request

Support

For issues and questions:

• Create an issue on GitHub
• Check the examples directory for usage patterns
• Review the API documentation

Note: This framework is designed for legitimate automation and testing purposes. Please respect website terms of service and implement appropriate rate limiting and error handling in production use.